antialias/haunt-fm

Go to file

Thomas Hallock 6a2be93556 Fix CLAP embedding extraction for transformers 5.x

In transformers 5.x, ClapModel.get_audio_features() returns a
BaseModelOutputWithPooling instead of a raw tensor. The 512-dim
embedding is in .pooler_output[0], not directly indexed. Added
backward-compatible extraction with hasattr check.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

2026-02-22 10:55:38 -06:00

Fix SQLAlchemy import: Real -> REAL

2026-02-22 08:59:10 -06:00

Initial haunt-fm implementation

2026-02-22 08:36:36 -06:00

Fix CLAP embedding extraction for transformers 5.x

2026-02-22 10:55:38 -06:00

.env.example

Initial haunt-fm implementation

2026-02-22 08:36:36 -06:00

.gitignore

Initial commit

2026-02-22 14:04:01 +00:00

alembic.ini

Initial haunt-fm implementation

2026-02-22 08:36:36 -06:00

CLAUDE.md

Initial haunt-fm implementation

2026-02-22 08:36:36 -06:00

docker-compose.yml

Initial haunt-fm implementation

2026-02-22 08:36:36 -06:00

Dockerfile

Fix Dockerfile: copy source before pip install

2026-02-22 08:51:45 -06:00

pyproject.toml

Initial haunt-fm implementation

2026-02-22 08:36:36 -06:00

README.md

Initial commit

2026-02-22 14:04:01 +00:00

README.md

haunt-fm

Personal music recommendation service — captures listening history, discovers similar tracks via Last.fm, embeds audio with CLAP, generates playlists